Human body model generation method, device and equipment and computer readable storage medium

Abstract

The invention provides a human body model generation method, apparatus and device, and a computer readable storage medium. The method comprises the steps of obtaining human body posture information; based on a human body generation network of a trained human body generation model, a first human body model is generated according to the human body posture information, and the first human body model is used for indicating a human body model in an unworn state; based on a human body dressing generation network of the human body generation model, generating a second human body model according to the first human body model and a preset implicit vector, the second human body model being used for indicating a human body model in a dressing state; and outputting a second human body model corresponding to the human body posture information. The method and the device can be applied to virtual customer service scenes of businesses such as financial science and technology, medical health, old-age care and the like, the first human body model is generated according to the human body posture information, and then the second human body model is generated according to the first human body model and the preset implicit vector, so that the human body model generation accuracy and efficiency are improved.

Description

Technical Field

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a method, apparatus, device, and computer-readable storage medium for generating a human body model. Background Art

[0002] Nowadays, the application scenarios of human body three-dimensional reconstruction are relatively wide. For example, in virtual customer service in financial business processing or medical, health and elderly care business scenarios, virtual customer service characters are generated through human body three-dimensional reconstruction. However, the human body models generated by most current human body generation methods are human body models without clothes. Such human body models lack authenticity and ornamental value. Therefore, it is necessary to add clothes to the human body models without clothes. In related technologies, the generation of human body models wearing clothes is generally achieved through human body posture vectors and clothing corresponding vectors. However, this method is easily affected by clothing vectors when generating human body models, resulting in problems such as large human body posture deviation and low clothing restoration of the generated human body models, which lead to insufficient accuracy of the human body models, and the amount of data that needs to be processed is large. Summary of the Invention

[0003] The present application provides a human body model generation method, apparatus, device and computer-readable storage medium, aiming to improve the generation efficiency and restoration degree of clothed human body models, and reduce the amount of data processing required to generate clothed human body models.

[0004] In a first aspect, the present application provides a method for generating a human body model, the method comprising the following steps:

[0005] Obtain human body posture information;

[0006] Based on the trained human body generation network, a first human body model is generated according to the human body posture information, where the first human body model is used to indicate a human body model in an unclothed state;

[0007] A human clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, wherein the second human body model is used to indicate the human body model in a dressed state;

[0008] Output the second human body model corresponding to the human body posture information

[0009] In a second aspect, the present application further provides a human body model generation device, the human body model generation device comprising:

[0010] Information acquisition module, used to obtain human body posture information;

[0011] A first generation module is configured to generate a first human body model based on a human body generation network of a trained human body generation model and according to the human body posture information, wherein the first human body model is used to indicate a human body model in an unclothed state;

[0012] A second generation module is configured to generate a second human body model based on the human body clothing generation network of the human body generation model and the first human body model and a preset latent vector, wherein the second human body model is used to indicate the human body model in a dressed state;

[0013] The model output module is used to output a second human body model corresponding to the human body posture information.

[0014] In a third aspect, the present application also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the human body model generation method as described above are implemented.

[0015] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the human body model generation method as described above are implemented.

[0016] The present application provides a human body model generation method, apparatus, device, and computer-readable storage medium. The present application utilizes a human body generation network based on a trained human body generation model to construct and generate a first human body model based on human body posture information, and utilizes a human body clothing generation network based on the human body generation model to generate a second human body model based on the first human body model and a preset latent vector, thereby completing the generation of a dressed human body model. The present application avoids the problems of large posture deviations and low clothing restoration in human body models generated by using human body posture vectors and clothing vectors to generate human body models wearing clothing. This improves the accuracy of the generated human body model and reduces the amount of data processing, thereby improving the efficiency of human body model generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 A schematic diagram of a flow chart of a method for generating a human body model provided in an embodiment of the present application;

[0019] Figure 2aA schematic diagram of a human body model provided in one embodiment of the present application;

[0020] Figure 2b A schematic diagram of a human body model provided in another embodiment of the present application;

[0021] Figure 2c A schematic diagram of a human body model provided in yet another embodiment of the present application;

[0022] Figure 3 is a schematic diagram of a human body model generating device provided in one embodiment of the present application;

[0023] Figure 4 A schematic block diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0025] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0026] Embodiments of the present application provide a human body model generation method, apparatus, computer device, and computer-readable storage medium. The human body model generation method can be applied to a terminal device, such as a laptop or desktop computer. It can also be applied to a server, such as a cloud server or server cluster.

[0027] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features therein may be combined with each other.

[0028] Please refer to Figure 1 , Figure 1 A flowchart of a method for generating a human body model provided in an embodiment of the present application.

[0029] like Figure 1 As shown, the human body model generation method includes steps S101 to S104.

[0030] Step S101: Acquire human body posture information.

[0031] Exemplarily, human body posture information is used to indicate the current posture and / or movement of the human body. Postures include but are not limited to squatting, sitting, standing, lying, etc., and human body movements include but are not limited to raising hands, lifting legs, and other movements of limbs or other parts of the body. That is, human body posture information can represent a human body raising hands while sitting, or raising hands while standing, etc., and there is no limitation here.

[0032] In a specific implementation process, human body posture information is determined and obtained through user input information, or human body posture information is obtained through a target image containing a human body; for example, if the user input information includes keywords such as standing and holding both hands horizontally, the corresponding human body posture information can be determined and generated based on these keywords; for another example, based on obtaining a target image containing a human body, the human body key points of the human body in the target image are identified to determine the human body posture information based on the human body key points, wherein the human body key points include but are not limited to key points for indicating the human head and limbs, key points for indicating human joints, etc.; or the user wears a motion capture device to determine the human body posture information through the signal transmitted by the motion capture device.

[0033] In a specific implementation process, after obtaining the human body posture information, the human body posture information is vectorized to obtain the rotation matrix values ​​for indicating the key points of the human body joints, and then a corresponding first human body model is generated in the human body generation model based on the rotation matrix values. It should be understood that the generated first human body model does not include facial features, facial expressions, hairstyle, etc. The first human body model can only display human body posture, movement, etc. In other embodiments, the facial expressions or facial features of human body models generated by other models can also be used to complete the facial expressions or facial features of the generated first human body model, which is not limited here.

[0034] Step S102: Based on the trained human body generation network, a first human body model is generated according to the human body posture information. The first human body model is used to indicate a human body model in an unclothed state.

[0035] Exemplarily, after obtaining the human body posture information, the human body posture information is input into the human body generation model, so that the human body generation model generates a first human body model in the human body generation network according to the input human body posture information. It should be noted that the first human body model generated at this time is generated according to the human body posture information, and only contains information such as human body movements and / or postures, and does not contain information such as clothing. Therefore, the generated first human body model is a human body model of a human body in a naked state.

[0036] In some embodiments, the human body generation network based on the trained human body generation model generates a first human body model according to the human body posture information, including: the human body generation network based on the trained human body generation model determines the human body posture information according to at least two target images containing human bodies, so as to generate the first human body model according to the human body posture information.

[0037] In a specific implementation, a target image containing a human body is input into a human body generation model, so that human body posture information of the human body in the target image is determined in the human body generation network, thereby generating a corresponding first human body model based on the human body posture information. It should be noted that if at least two target images are input into the human body generation model, the human body generation model can determine the background and human body in the target image based on the input target image. For example, the background in the target image can be determined through image comparison to extract the human body from the target image, thereby improving the accuracy of obtaining human body posture information and, in turn, improving the accuracy of the first human body model generated based on the human body posture information. It should be understood that after multiple target images are input into the human body generation model, the human body generation model can generate a first human body model corresponding to each target image based on each target image; or when the human body postures in several target images are similar, the human body generation model can also determine the target image whose human body posture information similarity is greater than a preset posture information threshold as the image to be processed based on the human body posture information of the human body in each target image, so as to generate a first human body model based on the human body in the image to be processed, thereby realizing automatic identification of the human body posture for which the human body model needs to be generated. It should be understood that the number of first human body models generated at this time is less than the number of target images input into the human body generation model; it should be noted that those skilled in the art can adjust the generation rules of the human body generation model according to the actual number of human body models generated or the number of target images obtained, such as generating a corresponding first human body model for each target image, or generating a first human body model based on at least two target images, etc., which is not limited here.

[0038] Step S103: Based on the human clothing generation network of the human body generation model, a second human body model is generated according to the first human body model and a preset latent vector, where the second human body model is used to indicate the human body model in the dressed state.

[0039] Exemplarily, after generating a first human body model in the human body generation network, the human body generation network transmits the generated first human body model to the human body clothing generation network, so as to generate a second human body model in the human body clothing generation network based on the first human body model and the obtained preset latent vector. It should be noted that the generated second human body generation model is a model of the human body in a dressed state, that is, the display state of the second human body model also includes the state of clothing worn on the human body, including but not limited to the swinging state due to the hem of the clothes not being completely attached to the human body, or the outline of the pants being slightly larger than the outline of the human thigh, etc., so that the user can feel that the second human body model is a model of the human body in a dressed state.

[0040] In a specific implementation, preset latent vectors are used to generate clothing for the human body model. Clothing includes, but is not limited to, short-sleeved and long-sleeved clothing, shorts, long pants, short skirts, and long skirts. It should be understood that different types of clothing correspond to different preset latent vectors. Users can determine the corresponding preset latent vectors based on the clothing they need to generate, thereby generating a second human body model.

[0041] In some embodiments, the vector dimension of the preset latent vector is less than or equal to 512.

[0042] In the specific implementation process, the preset latent vector has a vector dimension of 512 and does not distinguish between human posture information. That is, the preset latent vector can be applied to human body models with any action posture. The preset latent vector has a low dimension, which makes it easy to introduce it into the diffusion generation model as a control condition for downstream tasks.

[0043] In some embodiments, the human body generation network based on the trained human body generation model generates a first human body model according to the human body posture information, including: based on the parameter prediction layer of the human body generation network, determining the human body posture prediction parameters and the human body clothing prediction parameters according to the human body posture information; based on the first model generation layer of the human body generation network, generating the first human body model according to the human body posture prediction parameters, and transmitting the first human body model and the human body clothing prediction parameters to the human body clothing generation network; the human body clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, including: based on the second model generation layer of the human body clothing generation network, generating the second human body model according to the first human body model, the preset latent vector and the human body clothing prediction parameters.

[0044] Exemplarily, the human body generation network includes a parameter prediction layer and a model generation layer, wherein the parameter prediction layer is used to predict human body posture prediction parameters and human body clothing prediction parameters based on the input human body posture information; it should be noted that the human body posture prediction parameters are used to guide the generation of the first human body model in the model generation layer, so that the movement posture of the human body model in the generated first human body model is basically the same as the human body movement posture indicated by the human body posture information; after generating the first human body model, the human body generation network will transmit the first human body model generated in the model generation layer and the human body clothing prediction parameters generated in the parameter prediction layer to the human body clothing generation network, so that the human body clothing generation network generates a second human body model based on the first human body model and human body clothing prediction parameters output from the human body generation network, and the input preset latent vector.

[0045] For example, by generating corresponding human posture prediction parameters based on the acquired human posture information in the parameter prediction layer, the human movement posture of the generated first human body model can be made more accurate, and by predicting human clothing prediction parameters through human posture information, clothing can be generated according to the human movement posture, thereby improving the fit between the generated clothing and the human body model, making the generated second human body model more realistic and accurate, and reducing the risk of model penetration.

[0046] In some embodiments, generating the second human body model based on the first human body model, the preset latent vector and the human body clothing prediction parameters includes: determining a human body clothing prediction vector corresponding to the human body clothing prediction parameters; determining a target latent vector in the preset latent vector whose vector similarity with the human body clothing prediction vector is less than or equal to a similarity threshold; and generating the second human body model based on the first human body model and the target latent vector.

[0047] In a specific implementation, within the human clothing generation network, human clothing prediction parameters are vectorized to obtain a human clothing prediction vector. A target latent vector whose vector similarity to the human clothing prediction vector is less than or equal to a similarity threshold is determined from the preset latent vectors. Accordingly, corresponding clothing is generated on the first human body model based on the target latent vector, thereby obtaining a second human body model. It should be understood that the preset latent vectors include multiple latent vectors indicating different clothing types. Different human postures are suited to different clothing types. The human clothing prediction parameters can be used to filter out unsuitable clothing, thereby improving the efficiency of generating the second human body model.

[0048] Exemplarily, the vector similarity between a preset latent vector and a predicted human clothing vector can be determined using a Euclidean distance value. Specifically, the Euclidean distance value between the predicted human clothing vector and each preset latent vector is determined, and the Euclidean distance value of each preset latent vector is determined as the vector similarity with the predicted human clothing vector. In a specific implementation, the similarity threshold is a user-preset value, or a threshold determined based on the Euclidean distance values ​​of multiple preset latent vectors. For example, the Euclidean distance values ​​of the preset latent vectors are arranged in descending order, and the median of the sequence is determined as the similarity threshold; or the average of the Euclidean distance values ​​of all preset latent vectors is determined, and the average is determined as the similarity threshold. This is not limited herein.

[0049] In some embodiments, generating a second human body model based on the first human body model and a preset latent vector includes: generating clothing for a first area of ​​the first human body model based on the first human body model and the first latent vector; generating clothing for a second area of ​​the first human body model based on the first human body model and the second latent vector, wherein the first area and the second area are adjacent.

[0050] Exemplarily, different latent vectors are used to generate clothing for different areas of a human body model, thereby improving the efficiency of clothing generation for the human body model. In a specific implementation, the first area includes the upper body area of ​​the human body model, and the second area includes the lower body area of ​​the human body model. If a vector mapping relationship exists between the first latent vector and the second latent vector, the first latent vector and the second latent vector are used to indicate clothing with a connection between the upper and lower parts, such as a dress and a suspender dress. If a vector mapping relationship does not exist between the first latent vector and the second latent vector, the clothing indicated by the first latent vector and the second latent vector has no obvious correspondence. In other embodiments, the first area may also include one half of the body area of ​​the first human body model, and the second area may include the other half of the body area of ​​the first human body model, wherein the half of the body area includes an upper limb and a lower limb. In still other embodiments, the first area may also include the limbs of the first human body model, and the second area includes the torso area of ​​the first human body model. It should be understood that the specific reference areas of the first area and the second area can be adjusted according to the needs of those skilled in the art and are not limited here.

[0051] See also Figure 2a 、 Figure 2b and Figure 2c , Figure 2a A schematic diagram of a human body model provided in one embodiment of the present application; Figure 2b A schematic diagram of a human body model provided in another embodiment of the present application; Figure 2c A schematic diagram of a human body model provided in yet another embodiment of the present application.

[0052] For example, Figure 2aThe first human body model in the unclothed state is shown, that is, after the first human body model is input into the human clothing generation network, the obtained second human body model can be, for example, Figure 2b As shown, Figure 2b The second human body model shown is a human body model with clothing on the lower body area of ​​the human body model; or after the first human body model is input into the human clothing generation network, the second human body model obtained can also be, for example Figure 2c As shown, Figure 2c The second human body model shown is a human body model fully clothed in clothing. It should be noted that the second human body model obtained above is merely an example. The second human body model obtained can also be a human body model wearing clothing only in the upper torso area, or a human body model wearing other types of clothing. Those skilled in the art can adjust the model parameters and / or latent vectors according to actual needs to generate the desired second human body model. The specific type of clothing worn by the second human body model is not limited here.

[0053] In some embodiments, the human body clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, including: a generator based on the human body clothing model generation network generates a plurality of second human body models to be identified according to the first human body model and the preset latent vector; a determiner based on the human body clothing generation network performs model authenticity determination on each of the second human body models to be identified, and determines the second human body model to be identified whose determination result is true as the second human body model.

[0054] Exemplarily, the generation network of the human clothing model also includes a generator and a judger, wherein the generator can generate multiple second human body models to be judged based on the first human body model and the input preset latent vector, and transmit the second human body models to be judged to the judger; the judger judges each second human body model to be judged. Specifically, in the judger, the authenticity of each second human body model to be judged is determined. For example, the second human body models to be judged include a, b, and c; after processing by the judger, the authenticity of the second human body model a to be judged is determined to be true, and the authenticity of the second human body models b and c to be judged is false, then the second human body model a to be judged is determined as the second human body model.

[0055] Step S104: outputting a second human body model corresponding to the human body posture information.

[0056] For example, after the human clothing generation network generates the second human body model, the second human body model can be output for display to the user. In a specific implementation, since the generated second human body model has not been rendered, the second human body model can also be input into a rendering model to render the second human body model, thereby obtaining a rendered human body model. This application does not limit whether the second human body model is further processed. Those skilled in the art can determine whether to further process the second human body model based on the actual needs of using the second human body model.

[0057] By generating a human body model, a second human body model wearing clothing can be generated based on human body posture information, so that the generated human body model is richer in texture and has more details than the generated human body model without clothing, thereby improving the adaptability of the human body model to downstream tasks.

[0058] For example, in the scenario of self-service business handling in the field of financial technology, the generated second human body model can serve as the image of an intelligent customer service character to assist users in completing financial business handling; for example, in the scenario of intelligent consultation in the field of medical health, the generated second human body model can also serve as the image of a doctor or triage staff to answer health questions asked by users; it should be understood that the second human body model preferred by the user can be generated based on the user's portrait, and because the second human body model has high generation efficiency and authenticity, it can reduce the user's waiting time and enhance the user's experience of using the virtual customer service character.

[0059] The human body model generation method provided in the above embodiment constructs and generates a first human body model according to human body posture information through a human body generation network of a trained human body generation model, and generates a second human body model according to the first human body model and a preset latent vector through a human body clothing generation network of the human body generation model, wherein the latent vector is used to indicate the vector corresponding to the clothing, so that the clothes can be worn on the generated second human body model. The present application first generates a corresponding human body model according to the human body posture information, and then uses the latent vector to complete the generation of clothing on the human body model, avoiding the problems of large posture deviation of the human body model and low clothing restoration degree caused by using human body posture vectors and clothing vectors to generate human body models wearing clothing, and when the human body model needs to be changed, there is no need to generate the human body model again, thereby reducing the data processing amount of the human body model change, thereby improving the efficiency of the human body model change.

[0060] See also Figure 3 , Figure 3 This is a schematic diagram of a human body model generation device provided in an embodiment of the present application. The human body model generation device can be configured in a server or a terminal to execute the aforementioned human body model generation method.

[0061] like Figure 3 As shown, the human body model generating device 100 includes: an information acquisition module 110 , a first generating module 120 , a second generating module 130 , and a model output module 140 .

[0062] The information acquisition module 110 is used to acquire human body posture information.

[0063] The first generating module 120 is configured to generate a first human body model based on the human body generating network of the trained human body generating model and according to the human body posture information, where the first human body model is used to indicate a human body model in an unclothed state.

[0064] The second generation module 130 is configured to generate a second human body model based on the human body clothing generation network of the human body generation model and the first human body model and a preset latent vector, wherein the second human body model is used to indicate the human body model in the dressed state.

[0065] The model output module 140 is configured to output a second human body model corresponding to the human body posture information.

[0066] Exemplarily, the second generating module 130 includes a first region generating submodule and a second region generating submodule.

[0067] The first region generating submodule is configured to generate clothing for a first region of the first human body model based on the first human body model and the first latent vector.

[0068] The second region generating submodule is configured to generate clothing for a second region of the first human body model based on the first human body model and the second latent vector, wherein the first region and the second region are adjacent to each other.

[0069] Exemplarily, the second generation module 130 further includes a generator submodule and a determiner submodule.

[0070] A generator submodule, configured to generate a plurality of second human body models to be determined based on the generator of the human clothing generation network and the first human body model and the preset latent vector;

[0071] The determiner submodule is configured to determine the authenticity of each of the second human body models to be determined based on the determiner of the human clothing generation network, and determine the second human body model to be determined that is authentic as the second human body model.

[0072] Exemplarily, the first generation module 120 includes a parameter prediction submodule, a first model generation submodule, and a second model generation submodule.

[0073] A parameter prediction submodule, configured to determine a human posture prediction parameter and a human clothing prediction parameter based on the human posture information and the parameter prediction layer of the human body generation network;

[0074] The first model generation submodule is used to generate the first human body model based on the first model generation layer of the human body generation network according to the human body posture prediction parameters, and transmit the first human body model and the human body clothing prediction parameters to the human body clothing generation network.

[0075] The second model generation submodule is used to generate the second human body model based on the second model generation layer of the human body clothing generation network, according to the first human body model, the preset latent vector and the human body clothing prediction parameter.

[0076] Exemplarily, the second model generation submodule further includes a prediction vector determination submodule, a vector similarity determination submodule and a second human body model generation submodule.

[0077] The prediction vector determination submodule is used to determine the human body clothing prediction vector corresponding to the human body clothing prediction parameter.

[0078] The vector similarity determination submodule is used to determine a target latent vector from the preset latent vectors, whose vector similarity with the human clothing prediction vector is less than or equal to a similarity threshold.

[0079] The second human body model generation submodule is used to generate the second human body model according to the first human body model and the target latent vector.

[0080] Exemplarily, the first generation module 120 is further configured to determine the human body posture information based on the trained human body generation network according to at least two target images containing human bodies, so as to generate a first human body model according to the human body posture information.

[0081] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0082] The method of the present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0083] See also Figure 4 , Figure 4 This is a schematic block diagram of the structure of a computer device provided in one embodiment of the present application. The computer device may be a server or a terminal.

[0084] like Figure 4 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a storage medium and an internal memory.

[0085] The storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any one of the human body model generation methods.

[0086] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0087] The internal memory provides an environment for the operation of the computer program in the storage medium. When the computer program is executed by the processor, the processor can execute any one of the human body model generation methods.

[0088] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0089] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0090] In one embodiment, the processor is configured to execute a computer program stored in the memory to implement the following steps:

[0091] Obtain human body posture information;

[0092] Based on the human body generation network of the trained human body generation model, generating a first human body model according to the human body posture information, wherein the first human body model is used to indicate a human body model in an unclothed state;

[0093] A human clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, wherein the second human body model is used to indicate the human body model in a dressed state;

[0094] Output a second human body model corresponding to the human body posture information.

[0095] In one embodiment, when generating the second human body model based on the first human body model and the preset latent vector, the processor is configured to implement:

[0096] generating clothing for a first region of the first human body model according to the first human body model and the first latent vector;

[0097] Based on the first human body model and the second latent vector, clothing for a second region in the first human body model is generated, wherein the first region and the second region are adjacent to each other.

[0098] In one embodiment, when implementing the human clothing generation network based on the human body generation model and generating the second human body model according to the first human body model and the preset latent vector, the processor is configured to implement:

[0099] Based on the generator of the human clothing generation network, a plurality of second human body models to be determined are generated according to the first human body model and the preset latent vector;

[0100] Based on the determiner of the human clothing generation network, the authenticity of each of the second human body models to be determined is determined, and the second human body model to be determined that the determination result is true is determined as the second human body model.

[0101] In one embodiment, when the processor implements a human body generation network based on a trained human body generation model and generates a first human body model according to the human body posture information, it is configured to implement:

[0102] Based on the parameter prediction layer of the human body generation network, determining human body posture prediction parameters and human body clothing prediction parameters according to the human body posture information;

[0103] Based on the first model generation layer of the human body generation network, generating the first human body model according to the human body posture prediction parameters, and transmitting the first human body model and the human body clothing prediction parameters to the human body clothing generation network;

[0104] The processor implements a human clothing generation network based on the human body generation model, generates a second human body model according to the first human body model and a preset latent vector, and is further configured to implement:

[0105] Based on the second model generation layer of the human body clothing generation network, the second human body model is generated according to the first human body model, the preset latent vector and the human body clothing prediction parameter.

[0106] In one embodiment, when generating the second human body model based on the first human body model, the preset latent vector, and the human body clothing prediction parameter, the processor is configured to implement:

[0107] Determining a human clothing prediction vector corresponding to the human clothing prediction parameter;

[0108] Determining a target latent vector from the preset latent vectors, the target latent vector having a vector similarity with the human clothing prediction vector being less than or equal to a similarity threshold;

[0109] The second human body model is generated according to the first human body model and the target latent vector.

[0110] In one embodiment, when the processor implements a human body generation network based on a trained human body generation model and generates a first human body model according to the human body posture information, it is configured to implement:

[0111] Based on a human body generation network of a trained human body generation model, the human body posture information is determined according to at least two target images containing human bodies, so as to generate a first human body model according to the human body posture information.

[0112] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above description of human body model generation can refer to the corresponding process in the aforementioned human body model generation control method embodiment, and will not be repeated here.

[0113] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the human body model generation method of the present application.

[0114] The computer-readable storage medium may be an internal storage unit of the computer device described in the aforementioned embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a SmartMedia Card (SMC), a Secure Digital (SD) card, a flash memory card, etc., equipped on the computer device.

[0115] It should be understood that the terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0116] It should also be understood that the term "and / or" used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements that are inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system that includes the element.

[0117] The serial numbers of the embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments. The above description is only a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for generating a human body model, characterized in that: include: Obtain human body posture information; Based on the trained human body generation network, a first human body model is generated according to the human body posture information, where the first human body model is used to indicate a human body model in an unclothed state; A human clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, wherein the second human body model is used to indicate the human body model in a dressed state; Output a second human body model corresponding to the human body posture information.

2. The method for generating a human body model according to claim 1, wherein: Generating a second human body model according to the first human body model and a preset latent vector includes: generating clothing for a first region of the first human body model according to the first human body model and the first latent vector; According to the first human body model and the second latent vector, clothing for a second region in the first human body model is generated, wherein the first region and the second region are adjacent to each other.

3. The method for generating a human body model according to claim 1, wherein: The human clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, including: Based on the generator of the human clothing generation network, a plurality of second human body models to be determined are generated according to the first human body model and the preset latent vector; Based on the determiner of the human clothing generation network, the authenticity of each of the second human body models to be determined is determined, and the second human body model to be determined that the determination result is true is determined as the second human body model.

4. The method for generating a human body model according to any one of claims 1 to 3, wherein: The human body generation network based on the trained human body generation model generates a first human body model according to the human body posture information, including: Based on the parameter prediction layer of the human body generation network, determining human body posture prediction parameters and human body clothing prediction parameters according to the human body posture information; Based on the first model generation layer of the human body generation network, generating the first human body model according to the human body posture prediction parameters, and transmitting the first human body model and the human body clothing prediction parameters to the human body clothing generation network; The human clothing generation network based on the human body generation model generates a second human body model according to the first human body model and a preset latent vector, including: Based on the second model generation layer of the human body clothing generation network, the second human body model is generated according to the first human body model, the preset latent vector and the human body clothing prediction parameter.

5. The method for generating a human body model according to claim 4, wherein: The step of generating the second human body model according to the first human body model, the preset latent vector, and the human body clothing prediction parameter includes: Determining a human clothing prediction vector corresponding to the human clothing prediction parameter; Determining a target latent vector from the preset latent vectors, the target latent vector having a vector similarity with the human clothing prediction vector being less than or equal to a similarity threshold; The second human body model is generated according to the first human body model and the target latent vector.

6. The method for generating a human body model according to any one of claims 1 to 3, wherein: The human body generation network based on the trained human body generation model generates a first human body model according to the human body posture information, including: Based on a human body generation network of a trained human body generation model, the human body posture information is determined according to at least two target images containing human bodies, so as to generate a first human body model according to the human body posture information.

7. The method for generating a human body model according to any one of claims 1 to 3, wherein: The vector dimension of the preset latent vector is less than or equal to 512.

8. A human body model generating device, characterized in that: The human body model generating device comprises: Information acquisition module, used to obtain human body posture information; A first generation module is configured to generate a first human body model based on a human body generation network of a trained human body generation model and according to the human body posture information, wherein the first human body model is used to indicate a human body model in an unclothed state; A second generation module is configured to generate a second human body model based on the human body clothing generation network of the human body generation model and the first human body model and a preset latent vector, wherein the second human body model is used to indicate the human body model in a dressed state; The model output module is used to output a second human body model corresponding to the human body posture information.

9. A computer device, characterized in that: The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the human body model generation method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the human body model generation method according to any one of claims 1 to 7 are implemented.

Images